3. What You Will Learn
• Discover genome sequencing fundamentals
• Learn different analysis stages of sequencing, how AI applies to
each, genome editing fundamentals, and AI's role
• Discuss questions and considerations in "optimizing" human health
4. Topics
• Why Genomics?
• What is Genomics and What Can We Do About It?
• Opportunity and Outlooks
5. Topics
• Why Genomics?
• World Wide Impact on Health Challenges
• What is Genomics and What Can We Do About It?
• Opportunity and Outlooks
6. • Gene mutations responsible for causing illnesses in:
• Chromosomal diseases (e.g. Down Syndrome)
• Single-gene disorders (e.g. Sickle-Cell Anaemia)
• Multifactorial disorders (e.g. Diabetes)
• Mitochondrial disorders (e.g. Leber’s or LHON)
• Genes play a role in:
• infectious (e.g. AIDS, tuberculosis)
• non commutable diseases (e.g. cancer)
World Wide Impact on Health Challenges
Why Genomics?
7. World Wide Impact on Health Challenges
• Ongoing genomics revolution promises
• Change how diseases are diagnosed, prevented, and treated
• Provide some of the most personalized and effective medical treatments
• Genomic information and technology has the potential to improve
healthcare outcomes, quality, and safety and reduce cost
Why Genomics?
10. Topics
• Why Genomics ?
• What is Genomics and What Can We Do About It?
• Genomics Fundamentals
• Genetics vs. Genomics
• Genome Sequencing and Analysis
• AI Application to Genome Sequencing
• Genome Editing
• Genome Editing in Brief and Technologies
• Genome Editing Efforts
• AI’s Role in Genome Editing
• Opportunity and Outlooks
18. Interpretation and Sense Making
Sequencing Technology: 3 Stages of Analysis
What is Genomics?
For the biologist: analyze gene pathways
For the chemist: uncover structure-function relationships
For the pharmacologist: a first step in the drug discovery process
19. Base Calling
read alignment
variant calling
Interpretation
Sense Making
Read Alignment
Variant Calling
Interpretation
Sense Making
Sequencing Technology: 3 Stages of Analysis
What is Genomics?
20. Base Calling
read alignment
variant calling
Interpretation
Sense Making
Read Alignment
Variant Calling
Interpretation
Sense Making
Illumina Real-Time Analysis (RTA) software
Swift on Illumina Solexa Sequencing Platform
DeepVariant (Google)
PrimateAI (Illumina)
Diploid MOON
BaseSpace® Variant Interprator
Application of AI in Sequencing Technology
What is Genomics?
21. https://github.com/google/deepvariant
AI in Variant Calling: DeepVariant
What is Genomics?
DeepVariant (Google)
• Developed by Google
Brain team and Verily:
• Based on the same neural
network for image recognition
• Challenge: Sequencer errors
confound variant calling
• Published: 2016/12
https://www.biorxiv.org/content/biorxiv/early/
2016/12/21/092890.full.pdf
22. AI in Variant Calling: PrimateAI
What is Genomics?
• Developed by Illumina
• Challenge: Many variants can be
detected, but only some are
pathogenic in human
• Compared common missense
variants in other primate species
• Predict which may by pathogenic
• Published: 2018/08
https://github.com/Illumina/PrimateAI
https://www.nature.com/articles/
s41588-018-0167-z
BatchNormalization +
Relu +
1D Convolution
BatchNormalization +
Relu +
1D Convolution
1D Convolution
1D Convolution
1D Convolution
Only Primate
Conservation
Only Mammal
Conservation
Only Vertebrate
Conservation
1D Convolution
1D Convolution
Reference Sequence
Alternative Sequence
99 Vertebrate
Conservation
99 Vertebrate
Conservation
Pretrained Solvent
Accessibility Layers
Pretrained Secondary
Structure Layers
Concat
( 51,20)
( 51,20)
( 51,20)
( 51,20)
( 51,20)
( 51,20)
( 51,20)
( 51,40)
( 1,20,40)
( 1,20,40)
( 1,20,40)
( 1,20,40)
( 1,20,40)
( 5,40,40) ( 5,40,40)
( 51,40)
( 51,40)
BatchNormalization +
Relu +
1D Convolution
6 Layers
BatchNormalization +
Relu+
1D Convolution
+
+
( 51,40)
BatchNormalization +
Relu +
1D Convolution
( 5,40,40)
( 5,40,40)
( 1,40,40)
( 51,40)
( 51,40)
( 51,40)
( 51,40)
( 51,1)
BatchNormalization +
Sigmoid+
1D Convolution
( 1,1,1)
( 51,1)
BatchNormalization +
Relu +
1D Convolution
( 5,80,40)
( 51,80)
+ + ( 51,40)
( 51,40)
( 51,40)
( 51,40)
( 51,40)
( 51,40) ( 51,40)
( 51,40)
( 51,40)
( 51,40)
( 51,40)
( 51,40) ( 51,40)
23. AI in Interpretation: Diploid Moon
What is Genomics?
• Developed by Diploid, Belgium
• Challenge: identify the one or two
mutations responsible for the patient’s
condition hidden amongst 40,000 variants
• Autonomously diagnoses rare diseases
from NGS data using AI, reduce analysis
time from days or weeks to minutes
• Founded in 2014
http://www.diploid.com/
24. Topics
• Why Genomics?
• What is Genomics and What Can We Do About It?
• Genomics Fundamentals
• Genetics vs. Genomics
• Genome Sequencing and Analysis
• AI Application to Genome Sequencing
• Genome Editing
• Genome Editing Technologies
• Genome Editing Efforts
• AI’s Role in Genome Editing
• Opportunity and Outlooks
25. • A type of genetic
engineering in which
DNA is inserted,
deleted, modified or
replaced at a precise
location within a genome
of a living organism
The Mechanisms
Genome Editing
Illustrations from yourgenome.org
Insertion
Deletion
Modification
26. The Technologies
Genome Editing
Info or Comparing
Metrics
Technologies
ssDNA-RecA-CPP
nucleoprotein filaments
CRISPR base-
editor system
CRISPR/Cas9 TALENS ZFNs
Full name
ssDNA (single strand DNA)-
RecA-CPP (cell-penetrating
peptide) nucleoprotein
mediated homology-directed
recombination (HDR)
CRISPR Base
Editors (CRISPR-
based base editor
system)
CRISPR/Cas9 {(Clustered
Regularly Interspaced Short
Palindromic Repeats
(CRISPR)-associated protein
9 nuclease (Cas9)}
TALENs
(Transcription
activator-like
effector
nucleases)
ZFNs (Zinc
finger
nucleases)
Technology
advancement
Next gen - bacterial
recombinase-based
technology
Fourth gen -
CRISPR-
deaminase based
technology
Third gen - nuclease-based
technology
Second gen -
nuclease-based
technology
First gen -
nuclease-based
technology
Avoids double-
stranded break (DSB)?
YES YES NO NO NO
Does not introduce
unintended mutations?
YES YES NO NO NO
Adapted From TutumGene 2018 Genome Editing Technology Comparison
27. Cell Types
Genome Editing
• Somatic cells
• All body cells in a multicellular organism that are not sperm or egg cells
• Germ(line) cells
• Cells responsible for reproduction - sperm or egg cells
• Stem cells
• Unspecialized cells with self-renewal capacity that can divide limitlessly to
produce new stem cells, and can differentiate to different cell types in the body
29. Somatic, Stem Cell, and Germline Editing
Genome Editing
• Somatic genome editing: more mature; precise and regulated
• Stem cell editing can be used to treat or prevent disease or condition
• Germline genome editing may be needed for some genetic diseases
• that manifest themselves in a systemic way
• that are in a single but widespread tissue
• that are in a tissue not easily accessible
“Someday we may consider it unethical not to use germline editing to
alleviate human suffering.” – from A Crack in Creation
30. AI Tools - Examples
Genome Editing
• inDelphi, a Dash (a Python
framework) application, helps
scientists predict the
outcomes of end-joining
• Impact: make the editing
process more predictable,
controllable and useful
Published in Nature 2018
31. AI Tools - Examples
Genome Editing
• UT Austin’s CHAMP, or Chip Hybridized Affinity Mapping Platform,
built on NGS chip, spots editing mistakes
• Applications example: rapidly test a CRISPR molecule across a person’s entire
genome to foresee other DNA segments it might interact with besides its target
• Microsoft’s Elevation AI uses machine learning to predict so-called
off-target effects when editing genes with the CRISPR system
• Applications example: modify cells to combat cancer or produce high-yielding
drought-tolerant crops (e.g. corn and wheat)
32. Topics
• Why Genomics?
• What is Genomics and What Can We Do About It?
• Opportunity and Outlooks
• Questions and Considerations in Genome Editing Usage
• Market Opportunities
• Players in the field and discussions
34. Market Opportunities
Genome Editing market: 2017 $3.19B, CAGR 14.5%, 2022 est $6.28B
Future potential revenue from genome editing applications in:
• Research
• Medical fields
• Healthcare industry
• Food industry
• Agriculture
• Patent licensing
Estimated to be $25B by 2030
Source: grandviewresearch.com
35. Investments mainly in the US, China, and Europe. Some examples:
• Helix (US): health company that focuses on personal genomics and
connects people with insights into their own DNA
• Oxford Nanopore (UK):develops nanopore-based electronic systems
for analyzing single molecules, including DNA, RNA and proteins.
• iCarbon X (China):a China-based artificial intelligence platform for
health data company
• Prenetics (HK): leading genetic testing and digital health company in
southeast asia
AI in Genomics Players
36. Your Learning
• Discover genome sequencing fundamentals
• Learn different analysis stages of sequencing, how AI applies to
each, genome editing fundamentals, and AI's role
• Discuss questions and considerations in "optimizing" human health